AI Research Intern at NewsBreak (NLP team)

• Explored model context protocols (MCPs) and MCP selection pipelines to expand the functional capabilities of LLMs, deploying features as a Kubernetes service
• Developed and deployed new features for NBot, NewsBreak’s AI chatbot, in Python – increased click-through rate by ~50%
• Investigated methods for increasing LLM human-likeness through memory

NLP Research Assistant with Stanford Empirical Security Research Group

• Supervised by Professor Zakir Durumeric and PhD Hans Hanley
• NLP research on the spread of online misinformation, emotion features for controversial social media post detection, multilingual stance detection, and narrative differentiation

Intern at Meta (Facebook) University – Engineering

• Summer internship in iOS development
• Created iOS course collaboration app in 6 weeks with Objective-C (details & demo video here)
  • Integrated OAuth login, Facebook SDK APIs, and probabilistic text-similarity search for personalized recommendations

Stanford Electrical Engineering Teaching Assistant (ENGR 40M)

• One of 3-4 undergrad TAs (13 total) for Stanford’s largest EE course (150-200 students)
• Led weekly 3-hour lab sections (5-12 students) – circuit-building and programming
• Homework grader/Gradescope manager, held weekly office hours, graded labs/exams

Leveraging Graph Neural Networks for Analyzing User-Bot Engagement on Twitter (Stanford CS 224W Project)

• Exploration of Heterogeneous Graph Attention Network (HAN), GCN, and GAN to perform node classification between humans and bots on Twitter, analyze behavior of bots among human users, and identify bot farms
• Achieve accuracy of 0.74 and F1 score of 0.68 using HAN w/ BERT embeddings and focal loss
• Link to Medium article

Coursily iOS App

• Capstone Project from Meta University Internship for Engineering
  • Users can post and answer questions within selected course
  • Significant integration of Facebook SDK – includes login through OAuth, dynamic fetching/caching of Facebook Group posts
  • Uses a probability-based text similarity algorithm to recommend viewed posts to users based on posts that have been searched before
  • Store data, such as viewed posts and course information, on Parse
  • Code: Link to Github
  • Video Demo:

Tall and True Tales: Thoroughly Tracking the Takes and Trajectories of News Narratives from Trustworthy and Worrisome Websites

Hans W. A. Hanley, Emily Ito Okabe, and Zakir Durumeric
USENIX Security ‘25

Cardinal Ballet Company – Financial Coordinator

• Handle all funding, payments, and reimbursements for performances and events

Cardinal Ballet Company – Marketing Coordinator

• Company webmaster (JavaScript, CSS, HTML) for announcements and dancer profiles
• Create and distribute show programs

Grand Pas de Deux from Cardinal Ballet Company's Nutcracker

3rd Odalisque Variation from Cardinal Ballet Company's Le Corsaire